Français 
TEEPTRAK
New Machine Learning platform
Our Machine Learning platform makes it possible to process large quantities of industrial information in seconds. We currently offer an anomaly detection algorithm that can be used by any operator in a factory (it's SO simple and therefore SO impressive). Later this year, we plan to offer an optimization algorithm to help shop floor teams find the best settings to maximize production, energy efficiency and quality. Our platform uses two types of algorithms: Anomaly detection and optimization.
What makes our solution revolutionary: anyone with basic Internet access and very limited knowledge can use our system. We have worked and innovated very hard to create an automated and user-friendly interface using the latest technologies.
To detect anomalies, we use: 1) Isolation Fores. It is an anomaly detection algorithm that is used to identify outliers in a dataset. 2) Matrix profile. It has two main components: a distance profile and a profile index. To detect regime changes we basically use the following methods a) EWMA: Exponentially Weighted Moving Average (EWMA) b) PELT Algorithm: Pruned Exact Linear Time, it is a method to find points change in time series data. c) Bayesian Online Change points: this is a new approach based on Bayesian theory. To optimize processes, we typically use probabilistic graphical models: These are statistical models that encode complex multivariate probability distributions using graphs.
